Exposing + Native ISO

[Bill Anderson]

Man, we have it good today, when using film there were so many variable to be tested and accounted for in order to establish repeatable and accurate results - it seemed never ending at time: No two meters agreed, no two camera shutters were the same, iso had to be established for every camera and meter and lens combo - chemical constituents varied, water temp and PH variables, agitation methods, chemical exhaustion rates, film emulsion batch variables, it seemed if you cocked your hat the wrong way it would all go to shit; and yet these demands tuned one up to such a fine degree we could nail it every time - had to, it was sink or swim. ( :

 

[Elsie N]

Nonsense. (No offense intended.) How many times did Red say (And I am paraphrasing) We recommend shooting the MX @800 ISO. Hell that was practically the first thing they said about it when it was released.Edit: and nobody fell off of that slope. ;-)

They really drove the 800 point home with videos like this one:

http://www.red.com/learn/interviews/greg-st-johns-epic-transition

We need another Schilowitz, if that is possible.

IIRC, Jim may have initially been the one who suggested that number. And I don't remember him saying that was a hard number but merely the one he shoots. Maybe Jarred could post his preferred number and ask (if he is on site) or call Jim and ask him what his Dragon ISO sweet spot number is. If it's good enough for Jim, it's good enough for me (maybe '-).

But just like Mark said he prefers shooting 640 on Dragon, he also said that 1200, again IIRC, would correlate to MX's 800. That's a pretty large span.

 

[Mark L. Pederson]

Man, we have it good today, when using film there were so many variable to be tested and accounted for in order to establish repeatable and accurate results - it seemed never ending at time: No two meters agreed, no two camera shutters were the same, iso had to be established for every camera and meter and lens combo - chemical constituents varied, water temp and PH variables, agitation methods, chemical exhaustion rates, film emulsion batch variables, it seemed if you cocked your hat the wrong way it would all go to shit; and yet these demands tuned one up to such a fine degree we could nail it every time - had to, it was sink or swim. ( :

Not to mention that freekin' MoviCam Compact with the mags that always seemed to jam at the worst time. I hated that damn camera.

 

[Patrick Tresch]

Monitoring Dragon

Monitoring Dragon

Hello,

I was shooting Dragon today and I was thinking it would be great if RED would implement a new monitoring path : REDCOLOR+/GAMMA+

+ would be something like a better lowlight curve crushing less the blacks in favor of a gamma bump and a stronger denoise in the HMDI/HDSDI/LCD outputs of the camera.
This is to obtain a similar result to what we can achieve in RCX with a full debayer, Denoise set to 6 and a slight curve (what other manufacturers do in their debayered signal).

This would conduct in a less conservative iso setting (usually done to protect from noise) to bump a bit more the shadows (where the Dragon perfrom incredibly well), and this would also naturally push the DP to keep more DR in the highlights.

The traffic lights are hard to clip, or do show clipping after the highlights "color rendition" is lost or diminished. You don't clip but you can't get them out of the "Roll off zone" where they blend gently into white and use them as part of the picture.

What do you think? Too risky?


Pat

 

[Mark L. Pederson]

Hello,

I was shooting Dragon today and I was thinking it would be great if RED would implement a new monitoring path : REDCOLOR+/GAMMA+

+ would be something like a better lowlight curve crushing less the blacks in favor of a gamma bump and a stronger denoise in the HMDI/HDSDI/LCD outputs of the camera.
This is to obtain a similar result to what we can achieve in RCX with a full debayer, Denoise set to 6 and a slight curve (what other manufacturers do in their debayered signal).

This would conduct in a less conservative iso setting (usually done to protect from noise) to bump a bit more the shadows (where the Dragon perfrom incredibly well), and this would also naturally push the DP to keep more DR in the highlights.

The traffic lights are hard to clip, or do show clipping after the highlights "color rendition" is lost or diminished. You don't clip but you can't get them out of the "Roll off zone" where they blend gently into white and use them as part of the picture.

What do you think? Too risky?


Pat

Sounds messy to me. Remember - you can alter parameters in the camera which non-destructively change the monitor path.

FLUT can be handy to adjust on camera if you have a "monitor happy" DP who tends to under expose I have done that trick more than a few times You can also make and load curves in camera.

 

[Adrian Jebef]

Well ... like I have said over and over - there is not really a NATIVE ISO.

I'm saying ...

THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING THERE IS NO BASE RATING

Phil disagrees with you...

Or in a graphical explanation:

I think the inquiring minds on this thread are searching for the holy grail of RED; what Phil refers to here as "Optimal Exposure is found near the Native ISO of the sensor"


Hmmm. So... what is the Native ISO of DRAGON? 2000? 640? There's a Stop and a Half difference between the two. Or, in real-world terms, blown highlights or noisy blacks.

 

[Patrick Tresch]

Sounds messy to me. Remember - you can alter parameters in the camera which non-destructively change the monitor path.

FLUT can be handy to adjust on camera if you have a "monitor happy" DP who tends to under expose I have done that trick more than a few times You can also make and load curves in camera.

Curves&Flut are good but denoise as seen in RCX and some Graeme Magic to unleash the -5 to -8.5 stops still present in the Dragon RAW files is another story. That signal could also be recorded on the HDSDI feed, giving those Prores/DNXHD happy people the best proxies files possible "out of the camera".

Pat

 

[Mark L. Pederson]

Phil disagrees with you...

Does he really?

Yo Phil ... what's the native ISO of the Dragon?

 

[Blair S. Paulsen]

IMO RED put the "default" ISO for MX at 800 so people would be less likely to clip. This could be described as using the "alternate monitor brightness control" like a railing a few feet back from the scenic view to help keep people from falling over the side. I also buy the notion that they wanted enough leeway to facilitate channel balancing across a wide range of color temperatures - in part to compensate for the native response of CMOS sensors, as well as redistribution of Bayer CFA data that is inherently 50% green/25% blue/25% red.

Since multiple tests have shown that the Dragon is better in the highlights with a smoother transition into the clip than M or MX it could be argued that RED no longer needs to use a conservative "default" ISO biased to influence exposure choices down the scale a bit. Based on this theory, plus notes from Mark and others, I plan to use 640 as my preferred starting point - but its only a starting point. Full stop.

IAC, the issue of people mis-undersatnding the impact (or lack of same) of the ISO setting on what is written into the RAW data set has not helped RED's rep - fair or not. Many cameras have gain stages (usually analog) that actually boost the signal being captured and likely add noise reduction in proportion to the push. Since this processing step happens internally, many shooters rely on what the monitor is telling them and don't know or care what happened upstream. On RED, the metadata modulated version of the image shown on the monitor (and some tools) is essentially just a peek at one particular way of "developing" the RAW data set into a RGB picture at a particular reference ISO.

So if the RAW data set is set (code values fixed) - AND - the RGB monitor image derived from it is a sample characterization of that data set per settings, the real question might be how to best give useful feedback. My personal preference is to keep ISO and WB (3200 or 5600 only) the same every day so when I make changes in lighting or exposure those changes track the way my eye expects them to - I find lots of ISO changes too confusing.

I used to use the RAW view quite a bit but rarely look at it anymore, as, hold on to your hats, the only images anyone will ever see are RGB derived ones. That said, having a tool like EXPOSURE CHECK that is tied directly to sensor voltages and displayed in the most useful way possible (gamma encoding designed for the application) is nice to have available. If you want an absolute reading of the linear light values, I suggest you follow the advice of several posters and test thoroughly. In the field I just want a simple "check", as the name implies.

As others have noted, testing is the crux of the biscuit for realizing the full potential of any imaging device/signal chain.

Cheers - #19

 

[Adrian Jebef]

BTW: I like Phil's graph but I believe it's a bit misleading in the way it shows "expanded" or "compressed" highlight/shadow with respect to the -2 and +2 stops of exposure. Looking at it, the graph seems to suggest that by opening or closing your exposure +/-2 Stops or changing your ISO you keep the same Dynamic Range end points but squeeze in or squeeze out more or less highlight/shadow detail. This is not how digital CMOS-based sensors work. They have a linear sensitivity to light unlike a film stock whose sensitivity is logarithmic. Scene exposure values become discretely coded into a digital camera's DR (ie: 0-1023 in a 10-bit scale), they do not bend or curve like film does. Technically, there is no knee or shoulder with a digital cinema camera. The exposure curve (gamma) is interpreted directly from the captured linear code values. Though the curve can be interpreted numerous ways via post transforms it has no bearing on the top and bottom end of scene exposure.

I prefer this graph on the relationship of a sensor's native Dynamic Range and ISO changes:

 

[Mark L. Pederson]

I for one, need to spend some quality time around those that came before me, with light meters and variable situations, to appreciate the tools we have now, and how far we've come.

You are a wise man Will.

watch this video -

 

[Phil Holland]

Phil disagrees with you...

Does he really?

Yo Phil ... what's the native ISO of the Dragon?

Heh. There is no real "Native ISO". The cake is a lie. There is such a thing as base sensor sensitivity however and that's not what we're looking for though it's extremely important.

Before I continue, everybody real quick read this:
http://www.red.com/learn/red-101/iso-speed-revisited

i'm explaining ISO the way Red is explaining it.


What many of us refer to as Native ISO is the balancing point and even distribution of dynamic range above and below the midtone (most measure it off of 18% Gray), though not the way most would likely desire to shoot. Meaning the approximate middle ISO rating.

It's actually a horrible term and I should remove it from this graphic and change it to Base ISO.

However, I'll stir the pot as I just read the tonal placement on the new calibration and new OLPF.

Get ready. I'm going to reveal the rabbit in the hat here.


If you are looking for a balanced distribution of dynamic range on both sides of the of the approximate middle gray shooting with Dragon you will want to shoot between ISO 640-1600. You will notice a 1.3 stop variance in there, I'll go further into this in the next paragraph, as it's not really important. Notice I'm not calling this Native ISO, but rather lets call it the Balanced ISO or heck the more apt Base ISO.

Now, the actual truth. Since all of the dynamic range is accessible "all the time" at ISO values within the recommended operating range (ISO 250-2000) you are just exposing and rating for your midtone or 18% Gray.

In the graphic I made if you make a note of each "quadrant", meaning the individual boxes outlined by white across the grayscale gradient (consider those each stops) you can see clearly how stops get placed above middle gray while also visualizing the compression and expansion of tones. When the tones are expanded in the shadows, and in particular if you are dipping into the noise floor at higher ISO values like ISO 3200, this can reveal image noise in the shadows and darks. In that graphic this is illustrated by being able to see banding in the darkest side of the gradient.

 

[Adrian Jebef]

Though I think using a light meter has its merits in digital cinematography I find it impossible to feel confident using one for exposure unless one knows the Native Sensor ISO of the camera he/she is exposing for. Mark, perhaps you can enlighten me on how one uses a light meter correctly when exposing for an ISO RANGE between 200-2000?

 

[Adrian Jebef]

Heh. There is no real "Native ISO". The cake is a lie. There is such a thing as base sensor sensitivity however and that's not what we're looking for though it's extremely important.

What many of us refer to as Native ISO is the balancing point and even distribution of dynamic range above and below the midtone (most measure it off of 18% Gray), though not the way most would likely desire to shoot. Meaning the approximate middle ISO rating.

It's actually a horrible term and I should remove it from this graphic and change it to Base ISO.

If you are looking for a balanced distribution of dynamic range on both sides of the of the approximate middle gray shooting with Dragon you will want to shoot between ISO 640-1600. You will notice a 1.3 stop variance in there, I'll go further into this in the next paragraph, as it's not really important. Notice I'm not calling this Native ISO, but rather lets call it the Balanced ISO or heck the more apt Base ISO.

For simplicity's sake a Sensor's Base Sensitivity, Native ISO, Base ISO, Balanced ISO, or simply ISO (as one would find on a can of 35mm film) is the same thing. In terms of "scene exposure" ISO is a camera's or film's mid exposure point where there is an EVEN distribution of usable stops above and below middle grey. This standard definition has been used for about 100 years. I'm still perpetually confused why RED prefers to define their cameras with an ISO Range (200-2000 w/ Dragon) instead of a nominal exposure mid-point.


Phil, in your opinion, what have you found to be DRAGON's mid-point ISO (ie: 8 over / 8 under)?

 

[Jacobo Martinez]

Though I think using a light meter has its merits in digital cinematography I find it impossible to feel confident using one for exposure unless one knows the Native Sensor ISO of the camera he/she is exposing for. Mark, perhaps you can enlighten me on how one uses a light meter correctly when exposing for an ISO RANGE between 200-2000?

Taking into account what was said in this thread, you use the meter to create a look. This means if you dial in your meter iso 2000 you will get a look that has more highlight stops then shadows. The DR is more distributed in the highs. It is a look that you are maybe looking for a given story. But under the hood what you are doing is under exposing to have more highlights, or maybe because you are going for a textured look. then you compensate in the metadata chain with iso. Am I right mark?

 

[Mark L. Pederson]

Mark, perhaps you can enlighten me on how one uses a light meter correctly when exposing for an ISO RANGE between 200-2000?

Sure Adrian - this is exactly what I did the first day I got a Dragon -

First of all I would get to know my "film stock" - which for this example is the Dragon sensor.

Default ISO by manufacturer (RED) = ISO 800. I'm was pretty confident based on my experience with MX that RED's default ISO would be slightly higher than what IMO was optimal for "most" situations. I liked MX at ISO 500. But I keep an open mind - it is - after all - a new film stock.

I do some simple tests - I start with 800 - so I set my light meter to 800 - I place a subject holding a color chart against a neutral background - I take an incident reading next to the face of the subject - I open up one stop for skin tones (caucasian) - and I shoot a shot. Then, I bracket - + and 1 to 4 stops over and under. Then I look at the R3D files in RCX - push and pull them. And I mentally process what I see - based on my subjective impressions of what I like as a look and my personal tolerance for noise.

I shoot some more "stress tests" trying to break the camera. How far can I push this guy? How far can I pull this guy? I'm amazed how well low ISOs look and work. I test high ISOs. They work as expected - I'm more blown away by low ISOs and color and dynamic range.

I pay attention to the in-camera exposure meters as well - I notice that it really feels like I have to open up MUCH more to clip. After bringing in clips into RCX - I notice that I want MORE information that the histogram shows me. I REALLY want the Exposure Tool in RAW view.

My initial reaction was that rating the ISO at 800 on Dragon appeared to me to be much more accurate than rating the MX to 800 with respect the light meter. Further - I personally liked where 18% gray landed with respect to dynamic range and noise with a bit more light hitting the sensor than the base exposure (ISO800) based on that exposure method - so - if suspected that 640 was what I wanted.

I re-rated light meter to 640. Shot more bracketed tests and reviewed in RCX - while checking Exposure Tool in RAW view. I felt very warm and fuzzy. So ... I knew ... for me personally ... 640 was my personal "base" ISO for a "non specific" project.

iI tested low light and found ISO 1280 more my speed than ISO 2000 - as I have a much lower tolerance for noise now that I look at 4K monitors and see everything. I test bright exteriors - I'm blown away by 250 and 320 compared to MX.


So ... what do I set the light meter to shoot my movie? Example:

Low Budget Horror movie. The film has some day exteriors - but mostly dark interiors - and lots of night exteriors.

What's the creative? I like high contrast. So ... I'm gonna want to lean to a lower ISO and higher contrast ratio - brighter lights. I have low budget. SHIT - can't afford tons of bright lights - need to compromise on the night exteriors a bit - shit.

I'd rate at ISO 400 for day exteriors, 640 for interiors, 1280 for exteriors.

So my Dragon is three film stocks for this project - or - one film stock rated three different ways. Could I shoot the whole film at ISO 640 - sure - but, I need all the help I can get on night exteriors while trying to keep the contrast up - and I like the look of low ISO bight, daylight exteriors.

 

[Bill Anderson]

Shoot a test, using a spot meter, in order to evaluate/sync it with your chosen iso, and establish the iso's clipping point and max black with detail, it is then easy to PLACE values up and down that scale with impunity - you will never fear clipping or losing valuable shadow detail ever again. The power of the reflected light meter lies in its ability to show us relationships between every subject value - and since we know it reads everything at mid gray (18%) we can then see where values fall on the scale, and alter their position to suit our needs - as long as we have tested and accurately established the scale, just like the chart Adrian posted.
Take the iso 800 sample on the chart, we can see that anything at/above 7.4 stops above mid gray is clipped, and it's easy to see which scene values fall in or out of that zone if one is using a spot (reflected light) meter.

 

[Adrian Jebef]

Sure Adrian -

I'd rate at ISO 400 for day exteriors, 640 for interiors, 1280 for exteriors.

OK. Interesting. What kind of meter do you prefer to use? Incident or Spot?

I'm interested in those of you who continue to use light meters w/ digital cinematography as I've basically all but tossed mine years ago. Looking back I don't think I've every worked with a DP who selectively changed a film stock's ISO rating dependent on a particular scene (INT/EXT/DAY/NIGHT). Maybe if said DP was pushing or pulling the stock via processing but that will actually affect the film's exposure response, something that doesn't happen with cameras like DRAGON.

 

[Mark L. Pederson]

OK. Interesting. What kind of meter do you prefer to use? Incident or Spot?

I'm interested in those of you who continue to use light meters w/ digital cinematography as I've basically all but tossed mine years ago. Looking back I don't think I've every worked with a DP who selectively changed a film stock's ISO rating dependent on a particular scene (INT/EXT/DAY/NIGHT). Maybe if said DP was pushing or pulling the stock via processing but that will actually affect the film's exposure response, something that doesn't happen with cameras like DRAGON.

I'm an incident guy - only because that's what I had in film school. That's what I "know". I could not afford a reflected meter.

I actually want to buy a new one - I think I'm gonna go Sekonic L-478D LiteMaster Pro - http://www.sekonic.com/products/l-478dr/overview.aspx

 

[Jacobo Martinez]

I'm an incident guy - only because that's what I had in film school. That's what I "know". I could not afford a reflected meter.

I actually want to buy a new one - I think I'm gonna go Sekonic L-478D LiteMaster Pro - http://www.sekonic.com/products/l-478dr/overview.aspx

Phil Holland has a sekonic L-478, maybe he can tell us his experience using it.